1 What is known of Arctic kelps?

1.1 Present

  • Evidence suggests that many Arctic coasts should support seaweed
  • In Canada, kelp has been reported and documented along Arctic and subarctic coastlines
  • However, baseline measures of the extent of kelp communities are missing in much of the region


1.2 Future

  • Rapid environmental changes, such as declining sea ice, increased ocean temperatures, and freshwater inputs are occurring along Canadian coasts
  • Research suggests northern expansion of kelp forests with climate change
  • Therefore, the relationships between environmental factors and the presence of kelp forests in the Canadian Arctic are critical to understand


2 ArcticKelp project

  • This dive research conducted throughout the Canadian Arctic in 2014 - 2019
    • 5 - 20 m photograph quadrats

2.1 Campaigns


2.2 Mean cover


2.3 Question

  • Do the environmental drivers of kelp distribution differ for different functional groups?
    • Total kelp cover
    • Laminariales (Laminaria sp. + Sacharina sp.)
    • Agarum
    • Alaria

3 Environmental drivers

3.1 Abiotic data

  • NAPA (3-Oceans) model
    • Model outputs supplied by the Bedford Institute of Oceanography (BIO)

    • Based on the NEMO community ocean model (Madec and others, 2015)

    • Ice from the LIM3 model (Rousset et al., 2015; Vancoppenolle et al., 2009)

    • Daily surface resolution: 1998 to 2015
    • Five day (pentad) resolution at 75 depth layers
    • Tri-polar grid
      • 10 to 20 km resolution

3.2 Biotic data

  • Bio-ORACLE (Assis et al., 2018; Tyberghein et al., 2012)
    • Geophysical, biotic, and environmental variables
    • Collection from many different datasets
    • Surface and benthic coverage
    • Data from 2000 - 2014 for most
    • Single values per pixel; min, mean, max, and range for most
    • 5 arcdegree spatial resolution (~9.2 km at the equator)

4 Modelling distribution

  • Can we predict the % coverage of kelp groups?
  • Which variables are important?
  • What is the accuracy of the model?
  • How do models differ between groups?

4.1 Methods

  • Highly correlated variables were removed
  • The rest were fed to a random forest model (Breiman, 2001)
  • After many iterations the best variables were found
  • These best variables were used over many iterations again to find the best models

4.2 Variables

4.2.1 Total kelp

Data layer % Inc. MSE
Sea water temperature (mean at min depth) 92
Dissolved oxygen concentration (mean at min depth) 86
Ice divergence 80
Sea Water Y Velocity 0
kinetic energy 0
heat fluxes causing bottom ice melt 0

4.2.2 Laminariales

Data layer % Inc. MSE
Latitude 40
Longitude 27
Photosynthetically available radiation (mean) 23
Ice fraction 0
Chlorophyll concentration (mean at min depth) 0
Evap minus Precip over ocean 0

4.2.3 Agarum

Data layer % Inc. MSE
Ice thickness (cell average) 66
Light at bottom (mean at min depth) 53
Iron concentration (mean at min depth) 46
Net Downward Heat Flux 0
shear 0
total flux at ocean surface 0

4.2.4 Alaria

Data layer % Inc. MSE
total flux at ocean surface 8
non-solar heat flux at ocean surface 8
Sea Water Salinity 7
Light at bottom (mean at min depth) 0
kinetic energy 0
daily dynamic ice prod. 0

4.3 Confidence

4.3.1 Total cover


4.3.2 Laminariales


4.3.3 Agarum


4.3.4 Alaria


5 Results

  • Note that the colour scales are not the same between figures

5.1 Total cover


5.2 Laminariales


5.3 Agarum


5.4 Alaria


6 Conclusions

  • There should be quite a lot of kelp in the Arctic
  • There are different spatial projections for different groups
  • Alaria projections are likely incorrect and require more data
  • These projections provide a good platform for deciding future sampling locations

7 Further work

  • Better screening of variables used in model
  • More thorough model testing
  • Increase resolution of data
  • Introduce substrate data

8 Acknowledgements

  • Dr. Youyu Lu and Dr. Xianmin Hu for NAPA model access

  • This research was undertaken thanks in part to funding from the Canada First Research Excellence Fund, through the Ocean Frontier Institute.


References

Assis, J., Tyberghein, L., Bosch, S., Verbruggen, H., Serrão, E. A., and De Clerck, O. (2018). Bio-oracle v2. 0: Extending marine data layers for bioclimatic modelling. Global Ecology and Biogeography 27, 277–284.

Breiman, L. (2001). Random forests. Machine learning 45, 5–32.

Madec, G., and others (2015). NEMO ocean engine.

Rousset, C., Vancoppenolle, M., Madec, G., Fichefet, T., Flavoni, S., Barthélemy, A., et al. (2015). The louvain-la-neuve sea ice model lim3. 6: Global and regional capabilities.

Tyberghein, L., Verbruggen, H., Pauly, K., Troupin, C., Mineur, F., and De Clerck, O. (2012). Bio-oracle: A global environmental dataset for marine species distribution modelling. Global ecology and biogeography 21, 272–281.

Vancoppenolle, M., Fichefet, T., Goosse, H., Bouillon, S., Madec, G., and Maqueda, M. A. M. (2009). Simulating the mass balance and salinity of arctic and antarctic sea ice. 1. Model description and validation. Ocean Modelling 27, 33–53.